Autonomous Hacking Agent for Red Team
docker pull purpleailab/decepticon
docker run -d --name decepticon --restart always purpleailab/decepticon:latest
podman run -d --name decepticon --restart always purpleailab/decepticon:latest
skopeo inspect docker://docker.io/purpleailab/decepticon:latest
oras pull purpleailab/decepticon:latest
docker compose up -d
# Docker Compose override for development workflows.
#
# Auto-merge behaviour:
# - This file is merged with `docker-compose.yml` only when a Makefile
# target passes it explicitly (`-f docker-compose.dev.yml`); it is the
# contributor convenience layer โ `make dev`, `make cli-dev`, the
# benchmark harness pick it up via the COMPOSE_DEV/COMPOSE_WATCH vars.
# - The OSS installer (`scripts/install.sh`) does NOT download this
# file, so end users running the launcher (`decepticon start`) get
# only the base `docker-compose.yml` from `~/.decepticon/`.
#
# Skills now ship as package data under `decepticon/skills/` and are baked
# into the LANGGRAPH image via `COPY decepticon/` (resolved in-process by
# `decepticon/backends/__init__.py:SKILLS_LOCAL_PATH`). On a repo dev clone
# (`make dev`) skill edits are hot-reloaded by the `./decepticon` sync in
# `docker-compose.watch.yml` โ no dedicated `./skills` bind is needed (the
# old root `./skills/` directory no longer exists).
#
# Where to put YOUR overrides depends on which stack you are running:
# - **Launcher / OSS install** (`decepticon start`): drop a
# `docker-compose.override.yml` into `~/.decepticon/`.
# - **Repo dev clone** (`make dev`): add langgraph overrides below.
#
# See also: docs/proposals/156-compose-architecture.md.
services:
# No langgraph overrides currently needed โ skills hot-reload via the
# ./decepticon watch sync (docker-compose.watch.yml). Kept as a merge
# target so the COMPOSE_DEV/COMPOSE_WATCH `-f` chains stay valid and
# future dev-only overrides have a home.
langgraph: {}